RESUMO
BACKGROUND: For early screening and diagnosis of non-small cell lung cancer (NSCLC), a robust model based on plasma proteomics and metabolomics is required for accurate and accessible non-invasive detection. Here we aim to combine TMT-LC-MS/MS and machine-learning algorithms to establish models with high specificity and sensitivity, and summarize a generalized model building scheme. METHODS: TMT-LC-MS/MS was used to discover the differentially expressed proteins (DEPs) in the plasma of NSCLC patients. Plasma proteomics-guided metabolites were selected for clinical evaluation in 110 NSCLC patients who were going to receive therapies, 108 benign pulmonary diseases (BPD) patients, and 100 healthy controls (HC). The data were randomly split into training set and test set in a ratio of 80:20. Three supervised learning algorithms were applied to the training set for models fitting. The best performance models were evaluated with the test data set. RESULTS: Differential plasma proteomics and metabolic pathways analyses revealed that the majority of DEPs in NSCLC were enriched in the pathways of complement and coagulation cascades, cholesterol and bile acids metabolism. Moreover, 10 DEPs, 14 amino acids, 15 bile acids, as well as 6 classic tumor biomarkers in blood were quantified using clinically validated assays. Finally, we obtained a high-performance screening model using logistic regression algorithm with AUC of 0.96, sensitivity of 92%, and specificity of 89%, and a diagnostic model with AUC of 0.871, sensitivity of 86%, and specificity of 78%. In the test set, the screening model achieved accuracy of 90%, sensitivity of 91%, and specificity of 90%, and the diagnostic model achieved accuracy of 82%, sensitivity of 77%, and specificity of 86%. CONCLUSIONS: Integrated analysis of DEPs, amino acid, and bile acid features based on plasma proteomics-guided metabolite profiling, together with classical tumor biomarkers, provided a much more accurate detection model for screening and differential diagnosis of NSCLC. In addition, this new mathematical modeling based on plasma proteomics-guided metabolite profiling will be used for evaluation of therapeutic efficacy and long-term recurrence prediction of NSCLC.
RESUMO
Independent component analysis is commonly applied to functional magnetic resonance imaging (fMRI) data to extract independent components (ICs) representing functional brain networks. While ICA produces reliable group-level estimates, single-subject ICA often produces noisy results. Template ICA is a hierarchical ICA model using empirical population priors to produce more reliable subject-level estimates. However, this and other hierarchical ICA models assume unrealistically that subject effects are spatially independent. Here, we propose spatial template ICA (stICA), which incorporates spatial priors into the template ICA framework for greater estimation efficiency. Additionally, the joint posterior distribution can be used to identify brain regions engaged in each network using an excursions set approach. By leveraging spatial dependencies and avoiding massive multiple comparisons, stICA has high power to detect true effects. We derive an efficient expectation-maximization algorithm to obtain maximum likelihood estimates of the model parameters and posterior moments of the latent fields. Based on analysis of simulated data and fMRI data from the Human Connectome Project, we find that stICA produces estimates that are more accurate and reliable than benchmark approaches, and identifies larger and more reliable areas of engagement. The algorithm is computationally tractable, achieving convergence within 12 hours for whole-cortex fMRI analysis.
